Jointly Learned 3D Non-Cartesian Sampling With Wave Encoding and Reconstruction for Neurovascular Phase Contrast MRI

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging
• Fluid Dynamics

Background:

• Accelerating Magnetic Resonance Imaging (MRI) scans is crucial for reducing motion artifacts and improving patient comfort.
• 3D Phase Contrast (PC) MRI is essential for quantifying blood flow, but traditional methods are time-consuming.
• Wave encoding offers a promising strategy for accelerating MRI acquisition.

Purpose of the Study:

• To develop and validate an accelerated 3D PC MRI technique using jointly learned wave encoding and model-based reconstruction (MoDL).
• To compare the performance of the novel accelerated technique against conventional methods in terms of speed and accuracy.

Main Methods:

• A simulation framework and pseudo-fully sampled neurovascular 4D flow data were used to learn optimal wave encoding parameters and a MoDL network.
• Prospective 3D PC MRI scans were acquired in a flow phantom and 12 healthy participants using the learned wave encoding and a standard MoDL reconstruction.
• Acquisitions included the accelerated wave-encoded scan (2.25 min), a time-matched non-wave-encoded scan, a time-matched 3D radial scan, and a reference 3D radial scan (5.65 min).

Main Results:

• Learned wave scans in a phantom showed accurate flow rates and high correlation with reference scans (R² = 0.99).
• In vivo, wave-encoded scans exhibited reduced aliasing and blurring, with improved small vessel conspicuity compared to non-wave-encoded and time-matched radial scans.
• Quantitative flow analysis demonstrated similar or improved accuracy (coefficient of variation and intraclass correlation) for the accelerated wave scans compared to the reference scan.

Conclusions:

• Jointly learned wave encoding and MoDL reconstruction represent a feasible method for accelerating 3D PC MRI.
• The developed technique significantly enhances image quality and ensures accurate hemodynamic measurements.
• This acceleration holds potential for routine clinical application in neurovascular flow assessment.